Week 4 of the 2026 season produced 16 games and roughly a thousand offensive snaps, with per-team play counts that ranged from 32 to 86 and mostly clustered in the 50s and 60s. One week is a tiny sample, and nothing here outranks three weeks of prior-season evidence. But two results from the Sunday slate carried real statistical information, in opposite directions, and they're worth walking through on their own terms rather than by the final score alone.

The one play worth leading with

Atlanta beat New Orleans 45–24 on the road, and the box score undersells what happened on the ground. The Falcons ran the ball 33 times for 0.366 EPA/play — a positive rushing number on that much volume is unusual in any week, since rush EPA is normally dragged toward zero by sheer volume. Both backs produced it: Bijan Robinson carried 19 times for 145 yards at 0.339 EPA/play with 2 touchdowns, and Br. Robinson added 14 carries for 62 yards at 0.403 EPA/play with 3 touchdowns. The Falcons' passing game was even better — 0.507 EPA/play on 21 attempts — and across 54 total pass-or-rush plays Atlanta posted 0.421 EPA/play with zero turnovers and six offensive touchdowns. A 54-play game at that efficiency, with a run game doing real work alongside the pass game, is well outside the range of a typical NFL offensive performance.

A road win that matches its efficiency — and one that doesn't tell the story you'd guess

Indianapolis's 30–13 win in Washington was exactly as lopsided on the plays as it was on the scoreboard, just not for the reason the box score implies. Washington was favored by 4.5 points at home; Indianapolis won by 17. The per-play numbers back the margin up: Indianapolis ran 64 pass-or-rush plays at −0.086 EPA/play, split between a rough passing day (−0.371 EPA/play on 36 dropbacks) and an excellent one on the ground (0.281 EPA/play on 28 carries, with Jonathan Taylor going for 95 yards and 2 touchdowns on 20 of those carries). Washington was worse everywhere: 61 plays at −0.250 EPA/play, with the pass game at −0.242 on 39 attempts and the run game at −0.263 on 22 carries.

Indianapolis won a lopsided road game while being the team that coughed the ball up more. That's a signal about just how badly Indianapolis outplayed Washington on a per-snap basis, not a story about a clean, mistake-free win.

How the rest of the slate finished

The other 14 games, for the record: Dallas beat Houston 34–30, Green Bay beat Tampa Bay 17–14, Baltimore beat Tennessee 24–18, Kansas City beat Las Vegas 30–27, Seattle beat the Chargers 30–23, the Rams beat Philadelphia 24–20, Minnesota beat Miami 15–10, New England beat Buffalo 29–26, Chicago beat the Jets 23–12, Cleveland beat Pittsburgh 27–24, Jacksonville beat Cincinnati 22–17, Carolina beat Detroit 32–26, San Francisco beat Denver 24–14, and the Giants beat Arizona 36–24.

Two of those are worth a one-line flag rather than a deep dive, because the per-play numbers say they weren't upsets by this measure: the Rams actually outgained Philadelphia on a per-play basis too (0.059 EPA/play to −0.287), and New England's per-play edge over Buffalo (0.191 to 0.099) matched its final margin. Neither tells the same "score and process disagree" story as the two games above. The Giants-Cardinals score is a reminder that not every blowout is pure offense, though — part of New York's margin came on a 98-yard interception return for a touchdown, defensive scoring that a plain EPA-per-play number for the offense won't capture.

What would settle something as the season goes on

Does Atlanta's rushing split hold against a defense that isn't New Orleans's? Two backs combining for 33 carries at a combined 0.37-ish EPA/play is an extraordinary outcome for a single game; a following week at even half that rate would be the start of an actual trend rather than a one-week outlier.

Was Washington's performance about its offensive line and secondary, or Indianapolis's whole operation clicking at once? A team can lose by 17 with only one turnover and still be the better-coached side that just ran into a buzzsaw; only games against different opponents will tell you which was more true.

Everything above is roughly 50–65 plays per team, played once. Treat it accordingly as the 2026 season moves into Week 5.